概率模型驱动的无人机自适应多源信息轨迹估计OA
Probabilistic model-driven adaptive multi-source information trajectory estimation for unmanned aerial vehicles
精准的运动估计是无人机(UAVs)在复杂环境中实现自主飞行的关键,然而现有算法在特征退化或剧烈运动场景中难以维持高精度,甚至出现失效问题.为此,本文提出了一种前后端协同的多源信息融合轨迹估计算法.基于高斯过程,本文构建了一种平移-旋转运动解耦的连续时间轨迹模型.该算法利用运动信息先验与自适应平面协方差估计动态引导前端点云配准,有效抑制了复杂环境下的匹配漂移.此外,通过引入基于高斯过程(GP)置信传播的因子图优化框架,建立了前端配准与后端优化的双向信息反馈机制,提升了状态估计的全局一致性.通过数据集仿真及室外无人机实时飞行试验,验证了该算法在多种复杂场景中展现出良好的轨迹估计准确性及实时性.
Precise motion estimation is essential for the autonomous flight of unmanned aerial vehicles(UAVs)in complex environments.However,existing algorithms often suffer from accuracy degradation or even failure in feature-degraded or aggressive motion scenarios.To address this issue,this paper proposes a multi-source information fusion trajectory estimation algorithm with collaborative front-end and back-end.Specifically,a continuous-time trajectory model based on Gaussian processes(GP)is constructed,where translational and rotational motions are decoupled.By leveraging motion priors and adaptive plane covariance estimation,the algorithm dynamically guides front-end point cloud registration,effectively suppressing matching drift in complex environments.Furthermore,a factor graph optimization framework based on GP belief propagation is introduced to establish a bidirectional information feedback mechanism between front-end registration and back-end optimization.The global consistency of state estimation is enhanced.Comparative evaluations on datasets and outdoor flight experiments demonstrate that the proposed algorithm exhibits high trajectory estimation accuracy and real-time performance in various complex scenarios.
解伟豪;鲜斌
天津大学 电气自动化与信息工程学院,天津 300072天津大学 电气自动化与信息工程学院,天津 300072
无人机感知多源信息融合连续时间轨迹高斯过程置信传播
UAV perceptionmulti-source fusioncontinuous-time trajectoryGaussian processbelief propagation
《控制理论与应用》 2026 (9)
1881-1889,9
国家自然科学基金项目(62573313,91748121)资助. Supported by the National Natural Science Foundation of China(62573313,91748121).
评论